Overview
The engagement ran two parallel workstreams. The first built AI agents to improve the efficiency of notus’s existing service delivery. The second co-developed notusOS – a new client-facing platform built around AI agents to open up a second service tier with different economics.
Context
notus runs a fully managed service: clients provide raw input through interviews and conversations, and notus handles the full chain – strategy, content, publishing and audience growth. The client stays hands-off.
That model has clear strengths, but it carries a structural constraint common to high-touch service businesses: growth is tightly coupled to headcount. Every new client needs dedicated expert time across a multi-step process, so revenue and cost scale roughly in lockstep.
At the same time, the company had built up real intellectual property over years of client work – proprietary frameworks, playbooks and repeatable workflows. They were well-tested and effective, but they lived in people’s heads and internal documents. They powered the service, yet weren’t accessible as a standalone product.
That set up two workstreams: one to sharpen the efficiency of the existing model, one to package the methodology into a new, scalable offering.
AI-powered operational improvements
The goal here was to find where AI could take manual effort out of notus’s delivery without compromising the quality of the output.
The approach was targeted, not broad. We mapped the delivery chain end to end and pinpointed where human time went into tasks that could be partly or fully automated – turning raw client input into structured drafts, cutting back-and-forth in review cycles, and building internal utilities that gave the team more capacity without requiring prompt-engineering skills.
The principle throughout was augmentation: giving experienced people more leverage rather than replacing their judgement. The tools were built to fit existing workflows, not to impose new ones.
notusOS
The fully managed model serves clients who want to be completely hands-off. But there is a sizeable adjacent segment: people willing to be more involved. They want the methodology, the system and expert guidance – but they want to co-drive the work, not fully delegate it.
notusOS was conceived for them. It packages notus’s proprietary frameworks into a structured, client-facing platform where the client leads, supported by notus in an advisory role – a done-with-you model alongside the existing done-for-you business.
This is a service-to-product transition: taking the IP embedded in a high-touch service and making it accessible through a lower-touch, more scalable format. The hard part is deciding which parts of the methodology are self-serve, which need guided support, and which still require an expert – and redesigning how the company and client work together, since the roles shift significantly.
Waterglass contributed systems-architecture thinking, helped structure the platform for long-term extensibility, and advised on where to draw the line between product functionality and human expertise. notus owned the vision and product direction throughout.
The two workstreams fed each other. Mapping the delivery chain for automation surfaced what was codifiable and repeatable, which directly informed notusOS. And the two models serve different buyers at different price points – complementary, not competitive: the co-driven tier can be an entry point to the fully managed service, or an alternative for those who prefer to stay involved.
Outcomes
Delivery efficiency. AI tools deployed into the production workflow gave the team more capacity per person, cutting manual effort across key parts of the delivery chain.
New venture progress. notusOS moved from concept to working prototype – a client-facing platform for a second service tier, addressing a segment notus could not previously serve without proportional headcount growth.
Structured operations. Building both the tools and the platform forced a thorough audit of how the company actually works, surfacing undocumented processes and clarifying which parts of the methodology were genuinely differentiated.
Internal capability. The team built working fluency with AI tooling and began approaching the business through a product and systems lens – both of which carry forward beyond the engagement.
Engagement model
Waterglass was embedded in the notus team for the duration – part of day-to-day decisions, working alongside internal staff rather than as an outside contractor. On the AI tooling, we led technical design and implementation. On notusOS, the role was collaborative and supporting: contributing architecture and systems thinking while notus kept ownership of the product vision and direction.
Beyond the two core workstreams, the engagement included hiring support to build the internal roles needed to carry the work forward. As notus brought these capabilities in-house, our involvement phased out by design – the goal was to leave the team self-sufficient, not dependent.
